Somos um grupo de estudos de aprendizado de máquina e ciência de dados formado por alunos do @icmc_usp • We are a students group from Brazil focused in ML 🇧🇷
São Carlos, Brasil
Joined January 2020
- Tweets111
- Following96
- Followers408
- Likes37
data retweeted
Hoje às 18h estarei dando uma aula introdutória a Teoria da Informação aplicada a Machine Learning pelo grupo @data_icmc. Segue link: understandingdl.netlify.app/…
data retweeted
We are about to start our Understanding Deep Learning seminar! Today @stanislavfort will talk about Neural Network Loss Landscape in High Dimensions. Join us on Youtube: youtube.com/watch?v=stTzg8iU…
data retweeted
Had a great time giving a talk at @data_icmc today with Gal Kaplun on why "understanding generalization requires rethinking deep learning". The upcoming talks in this seminar look awesome understandingdl.netlify.app/…
See video of our talk here:
youtu.be/wi9mjnDfS7Y?t=370
data retweeted
A cool paper that uses @Love2Code's MiniGrid worlds toolkit to look at procedurally generating these worlds, and training agents inside of them based on curiosity-driven exploration: teapaw.org/robertarail/status/123…
Excited to share our @iclr_conf 2020 paper “RIDE: Rewarding Impact-Driven Exploration For Procedurally-Generated Environments”: arxiv.org/pdf/2002.12292.pdf w/ @_rockt at @facebookai! Code available at github.com/facebookresearch/…. 1/5
data retweeted
Understanding and Mitigating the Tradeoff Between Robustness and Accuracy
deepai.org/publication/under…
by Aditi Raghunathan et al. including @sangmichaelxie
#LinearRegression #Statistics
data retweeted
Understanding Self-Training for Gradual Domain Adaptation. Ananya Kumar, Tengyu Ma, and Percy Liang arxiv.org/abs/2002.11361
data retweeted
Some misc. thoughts on the societal implications of language model fine-tuning (thread):
1. As I’ve mentioned before, I think this is an instance of the “AI effect” - now that fine-tuning is the norm, people forget that it wasn’t always a thing, but it wasn’t, and it’s wild.
